With the rapid development of AI-generated video technology, its realism is increasing day by day, which also brings serious social security risks, such as the recent $25 million AI fraud case. To address this challenge, a research team at Columbia University School of Engineering developed an AI-generated video detection tool called DIVID (DIffusion-generated VIdeo Detector), which is designed to effectively identify and detect fake videos generated by advanced AI technology , providing a strong guarantee for maintaining network security and information authenticity.
DIVID’s core advantages are:
1. Targeting the new generation of AI: Specifically designed to detect videos generated by diffusion models, such as content generated by tools such as OpenAI’s Sora, Runway Gen-2, and Pika.
2. Technological innovation: DIRE (Diffusion Reconstruction Error) technology is adopted to identify AI-generated content by comparing the difference between the input video and the reconstruction output of the pre-trained diffusion model.
3. Built on success: DIVID is an extension of the text detection tool Raidar previously developed by the team, extending AI detection from the text field to the video field.
4. Openness: The research team has publicly released open source code and data sets, promoting the further development and application of this technology.
The emergence of DIVID marks an important advancement in AI content detection technology. With the widespread use of AI-generated videos in social media, journalism, and business, the importance of such detection tools has become increasingly apparent. It not only helps combat fraud and the spread of false information, but also provides new possibilities for maintaining authenticity and credibility in the digital world.
The open source release of DIVID provides a powerful weapon for the global fight against false videos generated by AI, which will help increase public awareness of AI technology and promote the construction of a safer and more reliable digital environment. It is believed that in the future, AI detection tools like DIVID will play an increasingly important role in maintaining network security and combating cybercrime.